Edge computing is particularly important for machine learning (ML) and other forms of artificial intelligence (AI) such as image recognition, speech analysis and large-scale use of sensors. Specific use cases may include video security surveillance, automated driving, connected industrial robots, and traffic flow and congestion prediction for smart cities. In the case of industrial IoT or self-driving cars, a processing delay between the device and the cloud could mean disaster. AI applications that are enabled by edge computing include access network analytics, video analytics, machine-to-machine analytics, augmented reality (AR) and location services.
A look at some of the best use cases of edge computing, and how they are supported by 5G technology…
Video analytics at the edge
Video analytics plays a significant role in various industries. For example, face recognition from traffic and security cameras helps in maintaining law and order. Several other types of analytics can be performed on video content such as object tracking, motion detection, event detection, flame and smoke detection, AI learning of patterns in live stream or video archives. Currently, video analytics is done on the cloud or on dedicated private servers depending on the need and functions to be established. Performing video analytics at the edge is both a requirement and an opportunity for several fields. Processing live video streams at the edge can improve surveillance and help in enforcing law and order. Two examples of this use case are face detection and incident identification and triggering, which would allow law enforcement officers to take immediate actions involving an incident.
Performing video analytics at the edge can support connected and self-driving cars. Live streaming of the scene as seen by a self-driving car needs to be analysed in a very short time to decide the actions to be taken by the car. A self-driving vehicle could already contain resources to process the scene instantaneously. Edge video analytics can process (or preprocess) farther scenes, and post-process video scenes for continual training and feedback. Similarly, video analytics at the edge is important for enabling smart cities. For example, traffic video analysis can be undertaken to route traffic in the most efficient way. Fire or smoke detection in an area can be identified instantaneously and ensure that the traffic is diverted from the danger zone by sending feedback to both the city infrastructure as well as to connected cars in the area.
Video analytics at the edge can also be used to enhance the real-life experience of event audiences such as sports, concerts and other shows. Videos from different camera angles at an event can be analysed and applied with AR/VR functions and presented to a live audience through large screens, smartphones and VR devices.
Content distribution networking and content caching at the edge
According to the Global Mobile Data Traffic Forecast update 2016-2021, global mobile data traffic grew 63 per cent in 2016 reaching 7.2 EB per month at the end of 2016. Videos accounted for about 60 per cent of mobile data traffic. By 2021, the global mobile data traffic is expected to reach 49 EB, with 78 per cent of the annual traffic expected to come from video.
As much of this traffic is video based, the possibility of redundant content being delivered to users in the same region is high. According to caching software provider Qwilt, over 80 per cent of the video traffic only consists of 10 per cent of the titles. Therefore, duplicates of the videos are being repeated, increasing the backhaul traffic and opex costs.
Most of the consumers are users of handheld devices such as smartphones, tablets and laptops. Therefore, the edge cloud becomes an appropriate infrastructure to cache content, which can significantly reduce the backhaul traffic. Considering that content caching may not be limited to videos, but extends to other data types such as music and documents, there is immense potential for telecom service providers to save on opex costs.
There are three primary methods of content caching – content caching based on traffic learning, targeted content caching and user-guided caching. Content caching has different workloads that may run together to perform the caching.
Some of them are content caching algorithms, data aggregators, ML codes, content traffic analytics and web servers.
Edge accelerated web
The edge accelerated web allows edge cloud services to be used by every smartphone user. Under most conditions, page load time is dominated by the front-end operations rather than the server in normal networks. The browser performs operations such as content evaluation and rendering. This consumes both time and power, which is essential for power-critical end devices such as mobile phones. By performing these operations at the edge, users can experience a quality browsing experience and save the battery power on their devices. Operations may include ad-block, rendering, content evaluation and video transcoding.
Speech analytics and derived workloads
Speech analytics comprises four components – speech recognition, machine translation, text-to-speech and natural language understanding. Major Tier I providers including Baidu, Microsoft, Google, Amazon, Apple and IBM offer application program interfaces that cover these areas. In addition, many big companies like Google, IBM, Microsoft and Baidu offer cloud-based speech analytics solutions. At the edge, there are device-level speech analytics solutions like Apple Siri or Amazon Alexa. With language user interface gaining ground as a more natural way of interfacing with the user, more speech analytics applications such as chatbots will continue to be developed in the foreseeable future.
Data processing at the edge for IoT
With IoT devices soon expected to produce trillions of GB of data daily, IoT is expected to be both the biggest producer and consumer of data. Billions of IoT devices will have a variety of uses, including smart city, smart retail, smart vehicles, smart homes and more.
Edge devices are, in theory, IoT devices. Video analytics and AR/VR will play an important part in the IoT ecosystem. For example, a face detection workload may be run for a device in a smart city setting, or for checkout in a smart retail shop, or as a part of AR for a private user. IoT workloads will also include all the AI workloads in terms of processing a data point.
One specific IoT-related workload is the IoT gateway. All IoT data is required to be processed differently at different latencies for varying purposes. Therefore, the computing capability to process all this data at different locations is necessary to fulfil the varying latency requirements. Thus, the edge cloud is an ideal location for data preprocessing (bidirectional), such as filtering and changing formats; data processing for latency-critical use cases and scenarios with connected components; and partial data processing and storage.
Data organisation and processing will be an important operation at the edge cloud. Fundamentally, data organisation entities range widely in complexity, from simple key-value stores that are designed for very fast data access to complex analytics operations.
Video surveillance and security applications
Multi-access edge computing (MEC) is used for analysing video streams from nearby surveillance IP cameras to conducting targeted searches in order to detect, recognise, count and track pedestrians, faces, vehicles, licence plates, abnormal events/ behaviours and other types of content in the video. Analysis and processing happen closer to the point of capture, thereby conserving video transmission bandwidth and reducing the amount of data routed through the core network. This new application will allow consumers to make payments for retail or entertainment purchases using MEC, 5G solutions and advanced facial recognition technology.
Connecting event attendees to video and VR applications
VR video streaming is another compelling MEC use case. People who attend large events, like conventions or sporting events, often struggle with basic access to data services via LTE or 4G networks. If basic data connectivity is not available, these mobile customers will be unable to conduct data-intensive tasks, such as video streaming or running VR-related applications. Network operators have proven that MEC can easily support high quality, data-intensive VR applications involving high resolution and 360-degree video.
Remote monitoring, network troubleshooting and virtual machines
AI is also applied to network operations at the edge. For example, CommSPs are using network analytics to monitor the behaviour of virtual machines. When issues or degradations are detected, network administrators can make quick decisions on how to handle them. Software-defined networking allows the distribution of network intelligence to the edge. With the ability to detect an issue or anomaly and address it quickly, rather than giving a delayed response in 10 or 20 minutes, edge helps in improving the quality of service significantly. s
Based on 5G Americas’ white paper, “5G at the Edge”